kwin_wayland coredumps are useless without prior manual intervention
We get bug reports right now that are un-actionable because kwin_wayland coredumps are too large and get truncated, making them un-debuggable when kwin_wayland crashes. Typically one of the following situations arises:
- User reports bug on kwin_wayland crashing; backtrace is useless because coredump is truncated
- User reports bug on some app or Plasma crashing; bug turns out to be caused by KWin crashing; bug triager asks reporter to get a kwin_wayland backtrace with coredumpctl; reporter obliges; backtrace is useless because coredump is truncated
In both cases, the user and bug triager alike end up frustrated, and the bug is not actionable, wasting everyone's time.
Apparently it is possible to fix this by adding ExternalSizeMax=6G to /etc/systemd/coredump.conf. But since this has to be done ahead of time, before the crash, it does not help people reporting an existing crash.
Is there a reason why kwin_wayland coredumps need to be so huge? They are much smaller on X11, so we don't run into this problem there.